OpenAI announced that ChatGPT users can now enable the "Lock Chat" feature for specific conversations; once locked, the conversation will be removed from the regular chat history, pinned lists, and general search results, moving to the "Items - Locked Chats" area, where a four-digit account PIN is required to reopen it; on supported mobile devices, users can also verify using Face ID, Touch ID, or other unlocking methods. This feature covers the free version, Go, Plus, Pro, Business, Enterprise, and Edu plans, while the Healthcare version is currently unsupported, indicating that OpenAI is gradually rolling it out to prevent sensitive conversations from being exposed during others' browsing.
The official guidelines also emphasize that locking a chat only changes the display location and opening method of the conversation, and does not equate to end-to-end encryption or the creation of a separate encrypted vault; while locked conversations will be securely stored in the account and users can access them later, information remembered from previous conversations will not be deleted due to locking, and memories prior to locking will not be overwritten by the new mechanism. This design aims to enhance "privacy control" in everyday usage scenarios, especially when phones or computers may be handed to others, preventing sensitive conversations from being directly displayed in the chat list.
Analysis of the Mechanism and Real-World Limitations of Conversation Locking
In practical operation, locked conversations will move to a dedicated memory space, where new messages can update the memory content of that area but will not be written into the regular conversation memory; simultaneously, files uploaded or generated within locked conversations will also receive the same protection. However, OpenAI has clearly stated that the locking feature does not change the overall retention period of chats and will not alter the company's data processing settings for "improving models for everyone." This means that users must still pay attention to account-level data governance and company policies during regular use.
From a technical and business perspective, this feature offers more nuanced privacy control, particularly when multiple users share the same device, helping to prevent temporarily exposed sensitive conversations from becoming a focal point for others; however, locking does not equate to end-to-end encryption and will not automatically delete overall data related to locked conversations, requiring users to manage their chat history and device security independently. OpenAI stated that locking only affects the display and opening methods while maintaining the existing data structure and basic settings for model training.
Practical Application Scenarios, Support Scope, and Future Outlook
From a practical standpoint, the locking feature is suitable for scenarios involving shared devices, such as in homes, offices, or educational environments; users can quickly isolate specific conversations from being viewed by others without disrupting the overall chat. As of the current version, this feature covers multiple subscription tiers, but the Healthcare version is not yet available, indicating that OpenAI is still evaluating the implementation costs and user needs across different versions. If more biometric integrations for unlocking are added in the future, it will help enhance the convenience and security of unlocking.
Additionally, reference background information shows that while locking conversations enhances perceived privacy, it does not mean that data is exempt from the influence of model training; OpenAI maintains its global policy of "improving models for everyone," simply providing clearer localized control options. Users concerned about data governance for highly sensitive content should also pay attention to account security policies, device unlocking settings, and the storage and access permissions for conversations and files.